Br J Hosp Med (Lond). 2026 May 22;87(5):54139. doi: 10.31083/BJHM54139.
ABSTRACT
AIMS/BACKGROUND: Malnutrition represents a major clinical concern in geriatric populations and is strongly associated with adverse outcomes and increased mortality, particularly among patients admitted to the intensive care unit (ICU). This study aimed to compare the predictive performance of the modified Nutrition Risk in the Critically Ill (mNUTRIC) score and the Prognostic Nutritional Index (PNI) for 30-day mortality in geriatric ICU patients.
METHODS: This retrospective study included 704 geriatric patients admitted to the ICU between December 2023 and June 2025 with a length of stay exceeding 24 hours. The mNUTRIC score and PNI were calculated using data obtained within the first 24 hours following ICU admission. Demographic and clinical characteristics were also evaluated and compared between groups according to these nutritional scoring systems. The primary outcome was 30-day mortality.
RESULTS: Of the 704 patients, 301 (42.76%) died during ICU follow-up, and 257 patients (36.51%) died within 30 days. Both the mNUTRIC score and PNI differed significantly between survivors and non-survivors (p < 0.001 for both). Patients with high mNUTRIC scores demonstrated significantly higher rates of both 30-day mortality and ICU mortality compared with those with low scores (p < 0.001 for all comparisons). Receiver operating characteristic (ROC) curve analysis showed that the mNUTRIC score had an area under the curve (AUC) of 0.753, with a cut-off value of 5.5, sensitivity of 0.743, and specificity of 0.638. In contrast, PNI demonstrated the lowest predictive performance among the evaluated scores, with an AUC of 0.624 and a cut-off value of 36.175.
CONCLUSION: Both the mNUTRIC score and PNI are useful for predicting 30-day mortality and clinical outcomes in geriatric ICU patients; however, the mNUTRIC score demonstrates superior predictive performance. Future research should further investigate their integration with established scoring systems and their utility in longitudinal trend monitoring.
PMID:42216593 | DOI:10.31083/BJHM54139
